BrowserOS neo: A Dedicated Browser for Running AI Agents on Your Machine
🌐 The open-source Agentic browser; alternative to ChatGPT Atlas, Perplexity Comet, Dia.
At a glance
- What is it?
- BrowserOS neo is an open-source Chromium-based browser for macOS and Windows that lets AI agents automate web tasks using your real logged-in accounts, with all session data stored locally.
- Who is it for?
- Engineers running Claude Code, Codex, or Cursor who need agents to act on logged-in accounts will find BrowserOS neo a direct fit: one-click agent connection, local session storage under ~/.browserclaw/, and a replay timeline for debugging what agents actually did. It is the wrong choice for CI pipelines and headless scraping, where Playwright or a dedicated agent-browser subprocess works better.
- Can I use it commercially?
- Yes, with strict conditions. AGPL-3.0 is a network copyleft licence: if people use a modified version over a network, for example as a hosted service, you must offer them its source code under the same licence.
- Is it still maintained?
- Yes. The repository received new commits within the last day.
- What is it written in?
- Mainly TypeScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 30, 2026, and from our analysis. They are not legal advice.
Editorial analysis
A Secondary Browser That Keeps Agents in Their Own Space
BrowserOS neo is not a replacement for Chrome or any daily browser. The README describes it as "a second browser, just for your AI agents," one that sits alongside Chrome and handles agent-initiated tasks in separate tabs while you keep using your existing browser normally. The core premise is that AI agents doing real work need access to your logged-in state: your inbox, your CRM, your LinkedIn feed. BrowserOS neo imports logins, bookmarks, and extensions from Chrome in a single click, then makes those credentials available to every agent that connects to it. Both BrowserOS neo and the related BrowserOS browser are Chromium forks, which means Chrome extensions are compatible with both. The repository hosts both products: BrowserOS neo (the browser agents drive) and BrowserOS (a Chromium fork with an AI assistant built into the new-tab page for human users). They run side by side.
How Agent Connections Work Through BrowserClaw
The agent-connection layer is called BrowserClaw, and its server is released separately from the main browser. The latest release at the time of the last repository push was claw-server/v0.0.58. BrowserOS neo discovers Claude Code, Codex, Cursor, VS Code, Zed, OpenCode, Hermes, OpenClaw, and Antigravity on your local machine automatically. Any AI that speaks MCP can connect. Once connected, agents get dedicated tabs. Multiple agents can run in parallel, each in its own tab, without interfering with each other or with any tab you have open yourself. The README states that for the same task, BrowserOS neo uses fewer tokens than alternatives such as Claude's Chrome extension or the Codex browser, though the README does not explain the mechanism behind that claim.
Installing BrowserOS neo on macOS and Windows
On macOS, the README provides a Homebrew path:
brew tap browseros-ai/tap
brew install --cask browseros-neoAfter the tap is registered and the cask installed, BrowserOS neo appears as a standard macOS application. For Windows, or as an alternative to Homebrew, the README links directly to installer downloads: a .dmg for macOS and a .exe installer for Windows at cdn.browseros.com/download/. There is no Linux build for BrowserOS neo; the README does not document one. After installation, importing from Chrome takes one click and brings over logins, bookmarks, and settings. Connecting an agent framework such as Claude Code also takes one click from inside the browser.
Live Dashboard and Session Replay
The new-tab page in BrowserOS neo acts as a live dashboard showing every agent currently running: which site it is on, what action it is performing, and how far along it is in the task. The documentation calls this the "cockpit" view, at docs.browseros.com/neo/cockpit. Every agent session is recorded to disk as a scrubbable video with a step-by-step action timeline. Sessions, screenshots, and history all go to ~/.browserclaw/ on the local machine. The README is explicit that this data never leaves the machine. The browser does send anonymous product-usage events (agent connect/disconnect, version, OS) to the BrowserOS team, but the README states it never sends URLs, page content, prompts, tool results, or screenshots. A single toggle in Settings disables even the anonymous events.
Where BrowserOS neo Is the Wrong Tool
BrowserOS neo is explicitly not designed for CI pipelines or test automation that needs a clean browser state. The README draws a direct contrast with Playwright and agent-browser setups, which spin up a fresh Chrome subprocess with no stored logins. That behaviour is correct for CI and web scraping that must not carry user credentials. BrowserOS neo imports and persists your real logins, which is what makes it useful for personal task automation and a liability in any context where credential isolation is required. Cloud browser services such as browser-use or Browserbase run in datacenters, which the README notes causes problems with sites like Twitter and LinkedIn that block datacenter IP addresses. BrowserOS neo avoids that because it runs on 127.0.0.1, but that also means it cannot scale across machines or run unattended on a server.
BrowserOS neo Against Atlas, Comet, and Dia
The README names Atlas, Comet, and Dia as alternatives that bundle a browser and an AI agent together. The key difference the README identifies: those products work only with their own AI. BrowserOS neo connects to the agents you already have, including Claude Code, Codex, Cursor, and others that speak MCP, so you are not switching AI providers to get agentic browser control. That is a meaningful practical difference for teams that have already standardized on a particular model or agent framework. The trade-off is that BrowserOS neo has no built-in AI of its own; it is purely an agent execution environment, and you need a separate AI tool to give it instructions.
Maintenance, Licensing, and What the Repository Contains
The last repository push was on 2026-09-24, and recent releases were published on 2026-09-22 and 2026-09-23, indicating active development. The repository is licensed under AGPL-3.0. The AGPL requires that any modified version of the software that you run as a service must also be made available as open source under AGPL; this is a meaningful constraint for teams planning to fork and modify the browser for internal use. The repository structure includes a packages/ directory, a tools/ directory, and a signatures/ directory alongside the documentation and GitHub workflows. Enterprise licensing inquiries are handled separately, as the README links to a founders@ email address for enterprise discussions.
Editorial conclusion
Engineers running Claude Code, Codex, or Cursor who need agents to act on logged-in accounts will find BrowserOS neo a direct fit: one-click agent connection, local session storage under ~/.browserclaw/, and a replay timeline for debugging what agents actually did. It is the wrong choice for CI pipelines and headless scraping, where Playwright or a dedicated agent-browser subprocess works better. Linux users are limited to BrowserOS, the human-facing Chromium fork in the same repository, since the README does not document Linux support for BrowserOS neo. Before deploying it to a team, check the MCP documentation at docs.browseros.com/neo/mcp to confirm your specific agent framework appears in the supported list.
Frequently asked questions
What is BrowserOS?
BrowserOS is a Chromium fork with an AI agent built into every new tab, designed for human users who want AI assistance while browsing. BrowserOS neo is a separate product from the same repository: a browser designed for AI agents to drive, not for human browsing.
Is BrowserOS free to use?
The README describes BrowserOS neo as free and open source. The source code is licensed under AGPL-3.0. The README does not document any paid tier for the desktop application.
How do I install BrowserOS neo?
On macOS, run brew tap browseros-ai/tap then brew install --cask browseros-neo. On Windows, download the installer from cdn.browseros.com/download/BrowserOS_neo_installer.exe. The README does not document a Linux build for BrowserOS neo.
How do I use BrowserOS neo with an AI agent?
After installation, BrowserOS neo automatically detects Claude Code, Codex, Cursor, VS Code, Zed, and other MCP-compatible agents running on your machine. Connecting any of them takes one click from inside the browser, after which you give the agent a task from the agent's own interface rather than from the browser.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/browseros-ai-browseros)